Thanks for the helpful information, but I didn't understand exactly what you were trying to say.
According to the above link, streaming in 4k/60fps from a non-standard camera faces two problems:
(1) lack of drivers and (2) hardware limitations.
1. Linux drivers for imx678
https://github.com/will127534/imx678-v4l2-driver
After a short search (I got lucky) I found a driver for this sensor written specifically for raspbian. Even if there are any problems with the driver, the groundwork has already been created and there is something to work with.
2. The hardware limitations are still a mystery to me. If I understood correctly, we need to pay attention to:
a) 4Lane (not 2) of CSI-2 processing capability;
b) CSI-2 protocol version;
c) a "powerful enough" ISP for processing raw data from camera;
d) a "powerful enough" CPU;
a) Lanes number.
IMX678 datasheet says: CSI-2 serial data output - 2 Lane / 4 Lane / 8Lane / 4Lane × 2ch
https://www.sony-semicon.com/files/62/f ... _Flyer.pdf
RPi5 and R CM4 have 4Lane of CSI-2
b) CSI-2 protocol version.
According to https://www.mipi.org/specifications/csi-2 and https://2384176.fs1.hubspotusercontent- ... r-2022.png we have:
Code:
CSI-2 ver. bitrate for D-PHY per Lane v1.1 1.5 Gbps v1.2 2.5 Gbps v1.3 2.5 Gbps v2.0 4.5 Gbps v2.1 4.5 Gbps v3.0 4.5 Gbps v4.0 9 GbpsCode:
for 10 bit1 frame = 3840×2160×10bit = 10.125 MB60fps = 60×10.125 = 607.5 MB/s = 4860 Mbit/s and for 12 bit1 frame = 3840×2160×12bit = 12.15 MB60fps = 60×12.15 = 729 MB/s = 5832 Mbit/sIn other words, even CSI-2 v1.1 is suitable for us (but CSI-2 v1.2 is better).
c) a "powerful enough" ISP for processing raw data from camera;
d) a "powerful enough" CPU;
I don't know how to calculate or measure ISP and CPU power for a certain bitrate of camera.
Statistics: Posted by abcdef — Fri Aug 01, 2025 2:11 pm